TLDR
- Anthropic introduced computer control functionality enabling Claude to operate Mac systems from afar
- Task delegation happens via mobile device; Claude manages applications, web navigation, and spreadsheet operations
- This capability integrates with Dispatch, enabling mobile-based task assignment to Claude
- The company cautions that this technology remains experimental and recommends keeping it away from sensitive information
- This development follows OpenClaw’s viral success, sparking an AI automation competition across the industry
A groundbreaking capability from [[LINK_START_1]]Anthropic[[LINK_END_1]] now allows Claude AI to assume control of users’ computers for autonomous task completion. The Monday announcement confirmed immediate availability for subscribers to Claude Pro and Claude Max plans.
You can now enable Claude to use your computer to complete tasks.
It opens your apps, navigates your browser, fills in spreadsheets—anything you'd do sitting at your desk.
Research preview in Claude Cowork and Claude Code, macOS only. pic.twitter.com/sVymgmtEMI
— Claude (@claudeai) March 23, 2026
Currently limited to MacOS devices, this innovation enables users to issue commands via smartphone while Claude executes operations on desktop machines — launching applications, conducting web searches, populating data sheets, and organizing files.
A demonstration video showcased a scenario where someone running behind schedule requested Claude to convert a presentation into PDF format and append it to a meeting invitation. The AI accomplished everything independently.
Integration with platforms including Google Calendar and Slack occurs through native connectors. When such integrations aren’t present, Claude can directly manipulate keyboard and mouse inputs to accomplish objectives.
Permission requests precede each new application access. Users maintain the ability to terminate operations at any moment.
How It Works With Dispatch
This computer control system operates in tandem with Dispatch, a recently launched component within Claude Cowork. Dispatch facilitates persistent communication with Claude across multiple devices, enabling assignment of recurring responsibilities.
When combined, these capabilities can manage automated morning summaries, execute software testing sequences, or monitor incoming messages without intervention.
[[LINK_START_1]]Anthropic[[LINK_END_1]] acknowledges that intricate operations might require multiple attempts. The designation as a research preview reflects the company’s intent to collect user insights and prioritize enhancement areas.
Security Risks and Warnings
Protective measures exist to minimize vulnerabilities, featuring automated detection of prompt injection exploits. Nevertheless, the organization advises against deploying this functionality with software containing confidential information — certain applications come pre-disabled for protection.
Cybersecurity professionals have raised concerns regarding autonomous AI agents. These systems can execute significant operations rapidly with minimal oversight. Additional dangers include malicious exploitation for unauthorized file or system access.
[[LINK_START_1]]Anthropic[[LINK_END_1]] pledged ongoing security enhancement as emerging risks surface.
Industry Competition Intensifies
This release positions [[LINK_START_1]]Anthropic[[LINK_END_1]] as a stronger competitor to OpenClaw, an open-source solution that gained widespread attention recently. OpenClaw enables task submission through WhatsApp or Telegram for local device execution. Nvidia recently introduced NemoClaw, an enterprise-focused iteration of this framework.
OpenAI entered this territory last month by recruiting Peter Steinberger, OpenClaw’s creator, to develop personal AI assistant technology.
For the present, Claude’s computer control remains available exclusively to MacOS users with Pro and Max subscriptions during its research preview phase.
